Minimum Time Ship Maneuvering using Neural Network and Nonlinear Model Predictive Compensator

在本研究中，提出了一种新的船舶最小时间操纵系统。本系统主要由两部分组成。一种是基于神经网络的最优解生成器，另一种是非线性模型预测补偿器。对于典型控制条件，神经网络通过插值预先计算的最小时间解来生成实际情况的最优解。在船舶复杂非线性动力学模型（MMG模型）的基础上，对各种最小时间机动的最优解进行了数值计算，并通过神经网络离线学习进行插值。此外，本文还利用计算最优解的非线性动力学模型，在线模拟了船舶的未来航向。在计算船舶未来航向的基础上，通过修改控制输入以实现最小时间机动来补偿由扰动引起的预测控制误差。首先，简要介绍了船舶最小时间操纵问题的求解方法和船舶动力学数学模型。其次，以最小时间平行偏差机动问题及其求解方法为例，介绍了该系统实现的可行性研究。然后，介绍了一种基于神经网络和非线性模型预测补偿器的最小时间机动系统。最后，对“盐次丸”（425总吨）训练舰进行了计算机模拟和在线试验。提出了一种基于神经网络和非线性模型预测补偿器的新型实用船舶最小时间操纵系统。在最小时间偏差问题中，系统在较短的计算时间内给出近似解，在实际情况下具有良好的跟踪性能。此外，实际海上试验证明了所提出系统的有效性。

In this research, a new minimum-time ship maneuvering system is developed.The proposed system is mainly composed of two parts. The one is a neural network based optimal solution generator and another is a nonlinear model predictive compensator. The neural network generates the optimal solution for real situation by interpolating pre-computed minimum-time solutions for typical control conditions. The optimal solutions for the various minimum time maneuvering are numerically computed based on the sophisticated nonlinear dynamical model of the ship (MMG model) and are learned off-line by the neural network for interpolation. Moreover, the same nonlinear dynamical model, which is used for the computation of the optimal solutions, is used to simulate the ship's future course on-line. Based on the computed ship's future course, the predictive control error caused by some disturbances is compensated by modifying the control input for minimum time maneuvering.First, the solving technique of the minimum-time maneuvering problems and the mathematical model of the ship's dynamics are briefly reviewed. Next, the minimum-time parallel deviation maneuvering problem and its solutions are introduced as an example of the feasible study realized by proposed system. Then, a minimum-time maneuvering system with neural network and nonlinear model predictive compensator is introduced. Finally, computer simulations and on-line experiments are carried out for a training ship Shioji Maru (425 gross tonnage).This research presented a new practical ship's minimum-time maneuvering system with neural network and nonlinear model predictive compensator. In the minimum-time deviation problems, the system gives approximate solutions in a short computing time and good tracking performance in real situations. Moreover, the actual sea trials demonstrate the effectiveness of the proposed system.